compoelem

Library for generating and comparing compositional elements from art historic images.

colabs to this project

https://drive.google.com/drive/folders/1ajWBqMry1zhEYhIC5dhzpnReOHeEJVuq?usp=sharing

Dataset

The dataset we used for the evaluation of our method can be found here: https://drive.google.com/file/d/1-lSW62XNsjwgeSSTtdlpMtAU9RjEqugD/view

Build and Setup:

git clone https://github.com/tilman/compoelem
cd compoelem
git submodule update --init --recursive
git submodule update --recursive

Build and Setup Openpose

Installation routine could change with an update of the git submodule. Tested with git commit b3e8abf from 20 Dec 2019. For an up to date installation procedure visit: https://github.com/tensorboy/pytorch_Realtime_Multi-Person_Pose_Estimation Depends on swig therfore run apt install swig or conda install swig for ubuntu or brew install swig for Mac OSX. Also make sure to download the model and place it under compoelem/compoelem/detect/openpose/pose_model.pth

conda install swig
cd compoelem/detect/openpose
pip install -r requirements.txt
cd lib/pafprocess; sh make.sh

Installation of the library:

python setup.py clean --all
python setup.py bdist_wheel
pip install dist/compoelem-0.1.0-py3-none-any.whl

or

python setup.py clean --all
python setup.py install

Usage:

See docs and tests for example usage.

Run Test Suite:

python setup.py pytest

Build the documentation:

first built the project and then run: pip install pdoc3 rm -rf docs && python setup.py clean --all && python setup.py install && pdoc --html --output-dir docs compoelem

Development: bump version and install locally

./release.sh major|minor|patch